Design of Experiments for Optimisation

Welcome to "Design of Experiments for Optimisation"!

Experimentation plays an important role in science, technology, product design and formulation, commercialization, and process improvement. A well-designed experiment is essential once the results and conclusions that can be drawn from the experiment depend on the way the data is collected.

This course will cover the basic concepts behind the Response Surface Methodology and Experimental Designs for maximising or minimising response variables.

This is not a beginner course, so to get the most of it, you need to be familiar with some basic concepts underlying the design of experiments, such as analysis of variance and factorial designs.

You can find it in my course “Design and Analysis of Experiments” or on several other courses and resources on the market.

The course starts with a basic introduction to linear regression models and how to build regression models to fit experimental data and check the model adequacy. The next section covers experimental designs for linear models and the use of central points to check the model’s linearity (lack-of-fit). By the end of the section, we will be using linear models to fit experiments with inaccurate levels in the design factors and missing observations.

By then, we will be ready for Response Surface Methodology. We will start with a factorial design to fit a linear model and find the path of the steepest ascent. And then, we are going to use a central composite design to fit a quadratic model and find the experimental conditions that maximise the response. Moreover, we will see how to analyse several responses simultaneously using two very illustrative and broad examples.

Finally, we will see how to use three-level designs: Box-Behnken and face-centred composite designs.

The whole learning process is illustrated with real examples from researchers in the industry and in the academy.

The analysis of the data will be performed using R-Studio. Although this is not an R course, even students that are not familiar with R can enrol in it. The R codes and the data files used in the course can be downloaded, the functions will be briefly explained, and the codes can be easily adapted to analyse the student’s own data.

However, if you are already familiar with using other DoE software, feel free to download the data and reproduce the analysis using the software of your choice. The results will be exactly the same.

Any person who performs experiments can benefit from this course, mainly researchers from the academy and the industry, Master and PhD students and engineers.